MaintNorm · DVC-Cluster Reproduction Package: Domain-Aware Dual-View Concept Contrastive Clustering for Industrial Maintenance Short Texts

Self-contained reproduction and archival package for a study on domain-aware dual-view concept contrastive clustering (DVC-Cluster) for industrial maintenance short texts. It contains the frozen MaintNorm corpus and knowledge resources, the method implementation (maintnorm_v8.py), the four-round experiment evidence chain (R1 main experiment, R2 fixed-K=50 supplement, R3 ID/metric unification, R4 frozen authoritative archive), the canonical result tables used by the manuscript, figure sources, and full SHA256 integrity manifests. Fixed protocol: K=50, ten pre-specified seeds, unified clean-test and external ID batches, and an independent three-expert Gold-240 semantic evaluation. The package documents a measured structure-semantic alignment trade-off and does not claim superiority over the general baselines (BGE, SimCSE, SCCL-BGE). Third-party material (MaintNorm corpus, base encoders) remains under its original license; see NOTICE.md.